P89 RNA-seq CAR transcript analysis

Overview

I compared coverage profiles and estimated abundances for CAR and related transcripts using three sets of samples:

• P89 bulk: sorted CD8+ T-cells, sequenced as bulk populations (expected to express CAR)
• P89 single-cell: CD8+ T-cells, captured using C1 instrument and sequenced invididually (expected to express CAR)
• P85 single-cell: MAIT-cells, captured using C1 and sequenced individually (not expected to express CAR)

 

Getting the data

Data presented below was generated and/or compiled from several sources:

• sample_metrics_data: sample annotation (e.g., donor ID, timepoint, etc.) as well sequencing & alignment metrics from RNA-seq processing
  
• sample_rapmap_data: read coverage measured across the length of the CAR transcript, based on mapping with the RapMap tool
• sample_salmon_data: abundance estimates (e.g., TPM, count) produced by the Salmon tool for CAR and several relevant transcripts when mapping to a modified human reference transcriptome (hg38)
• salmon_imgt_data: predicted/identified TCR junction sequences and alleles in single-cell libraries, as produced by assembly with Trinity followed by matching with IMGT High V-QUEST
• gff_file: custom-built GTF file describing where individual segments are located along the length of the CAR transcript

load("data/sample_metrics_data.RData")
load("data/sample_rapmap_data.RData")
load("data/sample_salmon_data.RData")
load("data/sample_tcr_data.RData")

gff_file <- "data/annotation/carPlus.gtf"
xcripts_gtf <- import.gff2(gff_file)

 

Results

Visualizing CAR coverage & abundance

The plots below show read coverage from RapMap mapping across the length of the CAR sequence. Segments in the transcript, corresponding to the gene parts used to build the construct, are depicted by colored boxes in each plot. Transparency (i.e., alpha) is scaled based on the estimated abundance of the CAR transcript (TPM) as measured by Salmon.

 

Bulk libraries from project P89

 

Single-cell libraries from project P89

 

Single-cell libraries from project P85

 

Inspecting trends in CAR coverage among single-cell P89 libs

To simplify the plots (and make it easier to distinguish between libraries), I’ll just plot the fitted line of coverage for each library.

 

CAR coverage vs. QC pass/fail

pass: median_cv_coverage < 1 AND mapped_reads_w_dups > 0.7

 

CAR coverage vs. presence of functional TCR junction

junction: functional TRA OR TRB junction sequence detected

 

CAR coverage vs. presence of TCR (paired junction)

TCR: paired TRA AND TRB for the same library

 

Inspecting abundance of non-CAR transcripts

The following human transcripts (which overlap the CAR sequence) were quantified by Salmon.

xcript_name	segment_version
CSF2	GMCSFRss_r1
CD28	CD28tm_r1
CD28	CD28tm_r2
CD28	CD28tm_r3
TNFRSF9	IgG4hinge_r1
CD247	CD3Zeta_r1
CD247	CD3Zeta_r2
EGFR	EGFRt_r1
EGFR	EGFRt_r2
EGFR	EGFRt_r3
EGFR	EGFRt_r4
EGFR	EGFRt_r5

 

Each plot shows CAR coverage across libraries, but colored based on the estimated abundance of the respective transcript.

 

Attempting to define CAR detection rules

I tried to come up with a relatively simple way to classify whether CAR was detected in a particular library.

 

Binarized CAR expression

expressed: log2(TPM +1) \(\gt\) 0 for CAR transcript

car_expr_tpm	nz_cov	n_libs
car_expr_tpm	FALSE	18
car_expr_tpm	TRUE	55
car_expr_tpm	NA	1
no_car_tpm	FALSE	444
no_car_tpm	TRUE	37
no_car_tpm	NA	4

 

Quantification-based rule

expressed: log2(TPM + 1) \(\geq\) 2.5 in CAR or EGFRt transcripts OR log2(TPM + 1) \(\gt\) 2 in all CAR or EGFRt transcripts

car_expr_quant	nz_cov	n_libs
car_expr_quant	FALSE	107
car_expr_quant	TRUE	58
car_expr_quant	NA	2
no_car_quant	FALSE	355
no_car_quant	TRUE	34
no_car_quant	NA	3

 

Coverage-based rule

expressed: \(\geq\) 10 positions with \(\gt\) 0 reads in ANY of CD19scFv, T2A, or EGFRt

car_expr_cov	nz_cov	n_libs
car_expr_cov	FALSE	2
car_expr_cov	TRUE	53
no_car_cov	FALSE	460
no_car_cov	TRUE	39
NA	NA	5

 

Revisiting trends

Plotting the remaining 53 libraries with CAR detected based on coverage.

 

Timepoint vs. donor

 

Functional junction detection vs. CAR detection

 

Paired TCR junction detection vs. CAR detection

 

Session info

## R version 3.2.1 (2015-06-18)
## Platform: x86_64-apple-darwin13.4.0 (64-bit)
## Running under: OS X 10.11.4 (unknown)
## 
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
## 
## attached base packages:
## [1] stats4    parallel  stats     graphics  grDevices utils     datasets 
## [8] methods   base     
## 
## other attached packages:
##  [1] cowplot_0.6.1        scales_0.4.0         ggthemes_3.0.2      
##  [4] rtracklayer_1.30.4   GenomicRanges_1.22.4 GenomeInfoDb_1.6.3  
##  [7] IRanges_2.4.8        S4Vectors_0.8.11     BiocGenerics_0.16.1 
## [10] viridis_0.3.4        ggplot2_2.1.0        dplyr_0.4.3         
## [13] tidyr_0.4.1          stringr_1.0.0        knitr_1.12.3        
## 
## loaded via a namespace (and not attached):
##  [1] Rcpp_0.12.4                highr_0.5.1               
##  [3] futile.logger_1.4.1        formatR_1.3               
##  [5] plyr_1.8.3                 XVector_0.10.0            
##  [7] futile.options_1.0.0       bitops_1.0-6              
##  [9] tools_3.2.1                zlibbioc_1.16.0           
## [11] digest_0.6.9               lattice_0.20-33           
## [13] nlme_3.1-126               evaluate_0.8.3            
## [15] gtable_0.2.0               mgcv_1.8-12               
## [17] Matrix_1.2-4               DBI_0.3.1                 
## [19] yaml_2.1.13                gridExtra_2.2.1           
## [21] Biostrings_2.38.4          grid_3.2.1                
## [23] Biobase_2.30.0             R6_2.1.2                  
## [25] XML_3.98-1.4               BiocParallel_1.4.3        
## [27] rmarkdown_0.9.5            reshape2_1.4.1            
## [29] lambda.r_1.1.7             magrittr_1.5              
## [31] GenomicAlignments_1.6.3    Rsamtools_1.22.0          
## [33] htmltools_0.3.5            SummarizedExperiment_1.0.2
## [35] assertthat_0.1             colorspace_1.2-6          
## [37] labeling_0.3               stringi_1.0-1             
## [39] lazyeval_0.1.10            RCurl_1.95-4.8            
## [41] munsell_0.4.3